The increasing pressure in industry to maintain tight control over processes has led to the development of many advanced control algorithms. Many of these algorithms are model-based control schemes, which require an accurate predictive model of the process to achieve good controller performance. Because of this, research in the fields of nonlinear process modeling and predictive control has advanced over the past several decades.;In this dissertation, a new method for identifying complicated block-oriented nonlinear models of processes will be proposed. This method is applied for LNL and LLN "sandwich" block-oriented models and will be shown to accurately predict process response behavior for a simulated continuous-stirred tank reactor (CSTR) and a pilot-scale distillation column. In addition, it will be shown to effectively model the pilot-scale distillation column using closed-loop, highly correlated input data.;Using the block-oriented models identified, a new feedforward control framework has been developed. This feedforward control framework represents the first that compensates for multiple input disturbances occurring simultaneously. Only a single process model is needed to account for all measured disturbances. In addition, it allows a plant engineer to develop the predictive model of the process from plant historical data instead of introducing a series of disturbances to the process to try to identify the model. This has the potential to considerably reduce the cost of implementing an advanced control scheme in terms of time, effort and money. The proposed feedforward control framework is tested on a simulated CSTR process in Chapter 4, and on a pilot-scale distillation column in Chapter 5.